Zero-Knowledge PII Redaction for Microsoft Word
Sanitize sensitive data across structured text, contextual NLP entities, and embedded image scans without losing Word formatting, tables, shading, or custom fonts.
Document Redaction Workspace
Supports Microsoft Word (.docx) packages up to 50MB
Matches SSN syntax, validates SSA area codes (no 000, 666, 900-999), and swaps with deterministic pseudonyms.
Rejects arbitrary 16-digit order numbers that fail Luhn. Valid cards are deterministically replaced.
Validates national and international area codes. Non-conforming numbers are ignored.
Full mailbox and domain validation. Substitutes realistic pseudonym email formatting.
Detects and validates IPv4 and IPv6 addresses, preserving valid subnet notation.
Distinguishes birthdates from invoice and contract dates using calendar layouts and age bounds.
spaCy NER model filtered to multi-token names to eliminate false positives on common solitary nouns.
Dual-signal matching: contextual NER combined with entity suffix patterns (Inc, LLC, Corp, GmbH).
Matches street number, name, unit/suite suffixes, and postal boundary heuristics.
Extracts ZIP media, runs OCR character recognition, and splices masked PNG/JPEGs back into the DOCX archive.